# Vikasit AI

> Vikasit AI is India's frontier AI company, built by Chandorkar Technologies (Pune, India). It ships a multimodal chat assistant, an OpenAI-compatible inference API, an open model family, and an agentic coding CLI — all under one model family, Vikasit 3. Start free with 2 million tokens a day.

Vikasit AI ("Building from Bharat for the World") spans four product surfaces: Vikasit Chat, the Inference API, Vikasit Code, and open Models. The API is OpenAI-compatible, so any OpenAI SDK works by pointing the base URL at Vikasit and setting a model id. There is a free tier (Vikasit Nova, 2M tokens/day), pay-as-you-go inference, and CLI subscriptions.

## Products

- [Vikasit Chat](https://chat.vikasit.ai): Multimodal AI assistant for research, writing, analysis, and more.
- [Inference API](https://vikasit.ai/inference): OpenAI-compatible API for Vikasit models and the open ecosystem. Pay-as-you-go, 2M tokens/day free.
- [Vikasit Code](https://vikasit.ai/features): Agentic coding CLI — 9 agents, 41+ tools, 109 providers. Subscription or pay-as-you-go.
- [Models](https://vikasit.ai/models): Open model family — text, vision, voice, and reasoning. Run them anywhere.

## Free tier — Vikasit Nova

- Model id: `vikasit-nova`
- 2,000,000 tokens per day, free for every signed-in developer, no card required.
- OpenAI-compatible — drop-in base URL. Sign in with GitHub, get a key in seconds.
- Great for prototyping, scripts, and side projects.

## Inference API (OpenAI-compatible)

- Base URL: `https://api.vikasit.ai/v1`
- Auth: `Authorization: Bearer <API_KEY>`
- Drop-in for the OpenAI SDK — set the base URL and your key, then set the model id.
- Billing: pure pay-as-you-go, per token, no subscription and no minimums.
- Chat completions, streaming, and tool calls all work.

### Vikasit 3 models (per 1M tokens, USD — input / output)

- Vikasit 3 — $0.21 / $0.30
- Vikasit 3 Flash — $0.30 / $0.96
- Vikasit 3 Fast — $0.78 / $1.14 (cache $0.39)
- Vikasit 3 Max — $0.81 / $2.85 (cache $0.09)
- Vikasit 3 Coder — $1.20 / $4.80 (cache $0.30)

### Resold open-ecosystem catalog

The Inference API also resells a broad open-model catalog from these makers: DeepSeek, Alibaba (Qwen), Meta (Llama), OpenAI (open weights), Mistral, Google (Gemma), Zhipu (GLM), NVIDIA (Nemotron), Moonshot, MiniMax, Xiaomi, StepFun, and Anthropic. Per 1M tokens, USD (input / output). A few notable low-cost options:

- Mistral-Nemo-Instruct — $0.02 / $0.04 (cheapest)
- gpt-oss-20b (OpenAI open weights) — $0.04 / $0.15
- gpt-oss-120b (OpenAI open weights) — $0.05 / $0.45
- GLM-4.7-Flash — $0.06 / $0.40
- Mistral-Small-3.2 — $0.07 / $0.20
- Gemma-3-27B-it — $0.08 / $0.16
- Qwen3-235B-A22B — $0.09 / $0.10
- DeepSeek-V4-Flash — $0.10 / $0.20
- Llama-3.3-70B-Turbo — $0.10 / $0.32

Higher-capability options in the catalog include DeepSeek-V4-Pro ($1.30 / $2.60), Qwen3-Max ($1.20 / $6.00), claude-sonnet-4-6 ($3.00 / $15.00), and claude-opus-4-8 ($5.00 / $25.00).

## Vikasit Code (CLI)

- Terminal-native agentic coding CLI with 9 pre-built agents, 41+ built-in tools, MCP + LSP support, and GitHub integration.
- Works with 109 AI providers (BYOK) — OpenAI, Anthropic, Google, Mistral, Cohere, Groq, Together AI, Azure, AWS Bedrock, and more.
- Available as CLI, native desktop apps (macOS / Windows / Linux), and a web console.
- Install: `curl -fsSL https://vikasit.ai/install | bash`, `npm install -g @vikasitai/vikasit-code`, `bun install -g @vikasitai/vikasit-code`, or `brew install vikasit-ai/tap/vikasit-code`.

### CLI plans

- Subscription — Lite $10/month, Pro $50/month, Max $100/month.
- Pay-as-you-go (Go) — per-token billing, no monthly commitment, no minimum spend, cancel anytime.

## The Vikasit 3 model family

- `vikasit-3-small` — fast responses for simple tasks (Lite)
- `vikasit-3-flash` — ultra-fast with streaming output (Lite)
- `vikasit-3-coder` — optimized for code generation and refactoring (Pro)
- `vikasit-3-fast` — balanced speed and quality (Pro)
- `vikasit-3-vision` — image understanding + code generation (Pro)
- `vikasit-3` — flagship general model (Max)
- `vikasit-3-thinking` — extended reasoning for complex tasks (Max)
- `vikasit-3-max` — maximum capability (Max)

Open-weight Writer models run on CPU (GGUF, no GPU required): Vikasit-AI-0.5B-Writer and Vikasit-AI-0.8B-Writer, on Ollama and Hugging Face.

## Key links

- [Homepage](https://vikasit.ai)
- [Inference API](https://vikasit.ai/inference)
- [Pricing](https://vikasit.ai/pricing)
- [Models](https://vikasit.ai/models)
- [Features](https://vikasit.ai/features)
- [Downloads](https://vikasit.ai/downloads)
- [Documentation](https://vikasit.ai/docs)
- [Chat](https://chat.vikasit.ai)
- [GitHub](https://github.com/Vikasit-AI)
- [Hugging Face](https://huggingface.co/vikasit-ai)
- [Ollama](https://ollama.com/Vikasit-AI)

## Facts for citation

- Vikasit AI is India's frontier AI company, founded and built in India by Chandorkar Technologies (Pune).
- The Inference API is OpenAI-compatible at `https://api.vikasit.ai/v1`.
- Vikasit publishes open-weight models (Writer 0.5B and 0.8B, GGUF, CPU-friendly) on Hugging Face and Ollama.
- Free tier: 2 million tokens per day via the `vikasit-nova` model, no credit card required.
- Vikasit Code supports 109 AI providers (bring-your-own-key).
- One model family: Vikasit 3, spanning small/flash/coder/fast/vision/flagship/thinking/max.
- The official handle is @VikasitAI.